xkcd: Fluid Speech
« previous post | next post »
Today's xkcd is (or should be) the illustration for a week or two in every introductory course on the sound side of language:
Mouseover text: "Thank you to linguist Gretchen McCulloch for teaching me about phonetic assimilation, and for teaching me that if you stand around in public reading texts from a linguist and murmuring example phrases to yourself, people will eventually ask if you're okay."
Randall Munroe probably already knows that "murmuring example phrases to yourself", though a useful exercise, is an imperfect approximation to the kinds of reductions that routinely happen in the fluent transformation from (phonological) symbols to (articulatory and acoustic) signals…
Unfortunately, a plurality of linguists (and the vast majority of psycholinguists) share (at least in their published work) the false belief that an accurate understanding of speech production and perception can be found by using recordings of subjects reading lists of de-contextualized sentences in a laboratory setting. (Though even there, speakers venture far from dictionary pronunciation fields…) Sociolinguists have of course championed the idea that you need to learn from patterns across a variety of speech styles and genres, including especially informal conversation — but the penetration of that idea across sub-disciplines has been surprisingly weak.
Here are a few posts where we've taken up adjacent topics:
"I'ma", 7/3/2005
"I'monna", 7/3/2005
"'On' time", 8/4/2005
"Finna and tryna", 8/5/2005
"I'ma stay with the youngsters", 5/14/2010
"The history of 'gonna'", 9/10/2010
"Ask Language Log: Writing 'gonna' or 'going to'", 6/25/2011
"Ima", 1/11/2012
"Gonna, gone, onna, a — on?", 8/10/2012
"Ask Language Log: 'Finna'", 11/4/2016
"From inflection to reduction", 1/10/2018
"On beyond the (International Phonetic) Alphabet", 4/19/2018
"Farther on beyond the IPA", 1/18/2020
"First novels", 3/13/2022
"Pronunciation evolution", 4/15/2022
"More post-IPA astronauts", 4/16/2022
For larger-scale discussions of (some of) the basic issues, see e.g. David Stampe's 1973 "A Dissertation on Natural Phonology"; Keith Johnson's 2004 paper "Massive reduction in conversational American English"; or my 2018 chapter "Towards Progress in Theories of Language Sound Structure".
Victor Mair said,
June 6, 2024 @ 6:58 am
"Welp, sup, yep, yup, nope" (5/29/20) — except that all of these end with a labial, far forward in the vocal tract.
Cervantes said,
June 6, 2024 @ 8:07 am
I gotta say, however, that computer processing of natural speech has gotten really good. I tried the freebie that comes with MS Word on an interview I recorded with a person whose dialect is quite different from mine, and it nailed it. I don't really have more to say other than it's interesting.
Mark Liberman said,
June 6, 2024 @ 8:19 am
@Cervantes: "I gotta say, however, that computer processing of natural speech has gotten really good. "
One key reason is that those systems' millions of hours of end-to-end training (via billions of dollars of GPU processing) no longer assume any particular dictionary pronunciations, but just adapt to whatever is in the training set…
Jarek Weckwerth said,
June 6, 2024 @ 8:21 am
Thank you, this is gold! Interesting to see ʌ in the third panel in view of the discussion we had on here some time ago…
Cervantes said,
June 6, 2024 @ 8:23 am
@ML — which is of course what our wetware does.
RICHARD R HERSHBERGER said,
June 6, 2024 @ 11:03 am
Geoff Lindsay has some enlightening YouTube videos where he takes a recording of someone speaking normally, then slices it up so you hear the sound that person actually made, not the sound you think he made based on the context.
Athel Cornish-Bowden said,
June 6, 2024 @ 11:19 am
The text under the cartoon certainly doesn't describe my speech. If I say "hot potato" in a sentence the first t (at the end of hot) may become a glottal stop (I'm not sure), but those in potato remain fully pronounced. (Perhaps I should add that I'm English and that I'm one of the few remaining speakers of the dying dialect known as RP.)
Scott P. said,
June 6, 2024 @ 11:26 am
The text under the cartoon certainly doesn't describe my speech. If I say "hot potato" in a sentence the first t (at the end of hot) may become a glottal stop (I'm not sure), but those in potato remain fully pronounced.
That's exactly what the mouseover text claims.
Philip Taylor said,
June 6, 2024 @ 12:48 pm
Although I cannot claim to be "one of the few remaining speakers of the dying dialect known as RP", I do endeavour to get as close to RP as I can when speaking in public, and I find that my treatment of the first "t" in "hot potato" varies with the placement of the latter in the sentence. In, for example, "He dropped the idea like a hot potato", my first "t" is fully pronounced; in, however, "Hot potatoes are selling like hot cakes", my first "t" is shorter but still nothing like a glottal stop.
Carol Saller said,
June 6, 2024 @ 1:10 pm
I can imagine a sentence for each of the pronunciations except the last:
Going to New York is easier than ever.
I'm goin' to the store – you want anything?
I'm gonna do that some day.
Could someone give me a sentence where the last pronunciation is likely? I can believe I use it, but can't think how. Thanks.
Lazar said,
June 6, 2024 @ 1:14 pm
Sadly why I find Forvo a bit less helpful than it should be. The individual words are generally given as "citation forms", and even the sentences tend to be given in a stilted, reading-from-a-book style. Trying to use it to suss out some nuance of real-life allophonics can be frustrating.
Ernie in Berkeley said,
June 6, 2024 @ 1:40 pm
> if you stand around in public reading texts from a linguist and murmuring example phrases to yourself, people will eventually ask if you're okay.<
Many years ago, my syntax professor told the story of being in a bar with another syntactician and bringing up example sentences for some very specific theory. Someone in the bar said to them, "What the hell are you talking about?"
Philip Taylor said,
June 6, 2024 @ 1:52 pm
Carol — "Could someone give me a sentence where the last pronunciation [I'm gonna do that some day] is likely ?" — It is not a pronunciation for which I would ever deliberately aim, but I can imagine that in an utterance as "Are you going to sit there all day, or I do need to kick your f*****g arse to get some work out of you ?" I might get closer than I would otherwise wish. But it would be /ˈɡɒn ə/, not /ˈɡʌn ə/.
Josh R said,
June 6, 2024 @ 7:17 pm
Carol,
I believe the last pronunciation might be rendered in eye-dialect as "gon'", pronounced like "gun", but with a schwa instead of the STRUT vowel, and with a less distinct nasal expression than "n" would suggest. I imagine that if it were transcribed by a (layman) native speaker, they would almost certainly parse it as "gonna".
Basically, "going to" in a sentence often finds itself unstressed and its constituent consonant clusters and vowels weakened to such a degree that only the initial consonant, the final (weakened) vowel, and a soupçon of nasalization remain.
Bob H said,
June 6, 2024 @ 8:24 pm
A few years ago I was on the line at an ice-cream stand behind an Italian tourist couple, who spoke English fairly well. They placed their order and the vendor asked [zæˈɾɪʔ]. They were completely stumped.
Julian said,
June 7, 2024 @ 2:56 am
Entering a quiet pizzeria in rural France. With my partner. Evening
The man turning to walk out nods to us and says, roughly: 'Swasieuda'. With a 'soupcon off nasalisation at the end.
Mark Liberman said,
June 7, 2024 @ 5:13 am
@Julian:
See "High vowel lenition/devoicing in French", 4/18/2024…
Benjamin E. Orsatti said,
June 7, 2024 @ 7:38 am
Bob's story rings true, especially where the Italian equivalent would have been "è tutto?", which is pronounced, in conversational (north-of-Napoli) Italian, /è tutto?/. Apart from those vowel-droppers south of the ankle, Italians like to pronounce all of their letters.
Bloix said,
June 7, 2024 @ 9:23 am
To keep myself occupied in my retirement, I tutor elementary school children. Typically these are students who are behind grade level, often including in reading, but have not been diagnosed with a specific learning or behavioral disability. Tutors are trained in school district teaching methods, including the use of phonics for reading instructs, and I believe that this is a correct choice and that phonics is certainly superior to the alternative "whole language" method (which as far as I can tell is more-or-less quackery). I find that by utilizing the phonics letter-by-letter "decoding" of text, children do make progress in reading, both aloud and to themselves. I also find that their teachers believe literally in phonics – that is, the teaching of a specific sound associate with each letter (with numerous exceptions, also taught) leads to the production of accurate speech sounds.
And yet, as this post show, it's well known that phonics-style rules don't come close to producing conventionally accurate speech sounds, which cannot be produced without applying a very large amount of information about context and meaning. And children, once they become moderately fluent readers beyond stage of "the fat cat sat on the mat," do produce conventional speech sounds while reading – that is, they don't follow the phonics rules either.
As I'm just a volunteer tutor, it's not my place to question what in any event seems to be a successful reading teaching technique, even if it appears to contradicts observed reality. But I wonder what linguists have to say about phonics, and about reading instruction generally.
VVOV said,
June 7, 2024 @ 10:16 am
> 'Swasieuda'
For the non Francophones in the comment section, what's this?
Philip Taylor said,
June 7, 2024 @ 10:36 am
VVOV — ['Swasieuda'] what's this ? — Hardly a Francophone, but I would postulate "Bonsoir, Monsieur dame !". I believe that Maurice Chevalier sang a song with a similar title but in the plural form.
Mark Young said,
June 7, 2024 @ 1:32 pm
@Carol: I think "Are you going to sit there all day?" is likely to have the fourth pronunciation. The more annoyed the speaker, the more likely the match, I suspect.
(Credit to Philip Taylor, who I suspect misunderstood your question, but gave a reasonable answer nonetheless.)
Chris Button said,
June 7, 2024 @ 2:30 pm
And the inputs are "tokenized" as numerical data.
Chas Belov said,
June 7, 2024 @ 5:31 pm
@Bob H: I too am completely stumped by [zæˈɾɪʔ].
@Carol Saller: I believe I pronounce the "o" in "gonna" in "I'm gonna do that some day." using the strut vowel, although I can't guarantee it's not a schwa as I have difficulty telling the two apart.
Doug said,
June 7, 2024 @ 8:06 pm
Chas Belov: " I too am completely stumped by [zæˈɾɪʔ]."
Presumably "Is that it?"
Julian said,
June 7, 2024 @ 11:26 pm
Bonsoir monsieur madame
Philip Taylor said,
June 8, 2024 @ 5:29 am
Chas — « I believe I pronounce the "o" in "gonna" in "I'm gonna do that some day." using the strut vowel, although I can't guarantee it's not a schwa as I have difficulty telling the two apart » — I thought of your comment when I had to tell the cat to stop clawing at my hand this morning — I definitely said /ʌ ʌ/, not /ə ə/.
/df said,
June 8, 2024 @ 6:45 am
Doesn't the t-ness of the /t/ in the "hot" of "hot potato" depend significantly on the speaker's choice of front or back /o/? RP-ish speakers will have their tongue forward when articulating the /o/ and are more likely to hit an (alveolar-)dental /t/ on the way to the subsequent plosive, whereas for "haht"-style speakers the tongue is rather in the middle of nowhere.
Philip Taylor said,
June 8, 2024 @ 7:28 am
That would seem an eminently sensible hypothesis, /df — my near-RP /pɔ ˈteɪ təʊ/ certainly seems to pre-dispose my mouth to formulating a clear /t/, whereas if I endeavour to simulate what I believe to be the <Am.E> pronunciation (/pɑ ˈdeɪ dəʊ/) the /t/ definitely moves towards a /d/.
Rodger C said,
June 8, 2024 @ 12:19 pm
Peter Taylor: My AmE pronunciation of "potato," which I think is usual, is /pə ˈteɪ dʌʊ/. The vowels may vary, but the /t/ isn't voiced before a stressed vowel.
The young C. S. and W. H. Lewis famously mocked their father for saying "p'dayta."
Andrew Usher said,
June 8, 2024 @ 4:27 pm
Per Jarek Weckwerth, the STRUT vowel in the third panel of the cartoon is surely a mistake – it doesn't match the picture, either. 'Gonna' when unstressed, and it almost always is, is pronounced /gənə/ in both American and British speech (it can be /gənu/ before a vowel, but that's a different thing). When stressed, /ʌ/ might be a possibility, though I certainly keep schwa; Philip Taylor's /gɒnə/ is probably common for Brits but it must be a spelling pronunciation (as the full form doesn't have that vowel or anything that would plausibly syncopate to it).
STRUT and schwa are really no less different in the standard variety of American as in the standard variety of British, and it's unfortunate that there's a legacy of teaching the contrary, perhaps aided by respelling systems, with no clearly better option, using 'uh' for both.
Regarding 'hot potato', it was asking about the t in 'hot', not the pronunciation of 'potato' – which, by the way, has schwa in the first syllable – I categorically would produce that t as unreleased /t/, which counts as 'fully pronounced' for me, though some people think it has to be exploded and aspirated – which is not natural finally even in isolation. I don't understand the remark of /df: the vowel in 'hot' is never front and in any case vowel quality is not a major influence on t-glottaling.
k_over_hbarc at yahoo dot com
Lazar said,
June 8, 2024 @ 7:58 pm
@Andrew Usher: I've noticed that some Brits have an innovative /ˈə/ (≠/ʌ/) that shows up in stressed tokens of gonna and because (you can hear this in Cambridge's online recording for British gonna, which sounds very different from gunner). In Southern and Black US English gonna takes /ɔː/, or in the broadest accents /oʊ/, and tends to reduce to *gon', also with /ɔː/ or /oʊ/.
On hot, yeah, I'm doubtful that vowel frontness has much of anything to do with a speaker's use of /t/ allophones. Hot can take a central [ä] in "NCVS" accents, although I don't think it reaches front of center except in parody.
Julian said,
June 8, 2024 @ 8:11 pm
For anyone interested, there's an extended riff on this theme in the works of Prof. Afferbeck Lauder (Australian, floruit 1960s), of which the best known is Let Stalk Strine:
"Dja getny btydas fer dinner?" = Did you get any potatoes for dinner?
"Doosa fiver mite wea?" = Do us a favour, mate, will you?'
The blurb of the reprint concludes:"it is essential reading frenny won inner rested ennair wire flife. Saul ear—tiger lookin star torgon Strine!"
Similarly earlier in C J Dennis's The Sentimental Bloke, which was a huge hit during World War 1.
https://en.wikipedia.org/wiki/The_Songs_of_a_Sentimental_Bloke
"The world 'as got me snouted jist a treat; Crool Forchin's dirty left 'as smote me soul" etc etc
Personal explanation: The Sentimental Bloke, being a "classic", I've tried it a few times but never got far. You have the uneasy feeling that with all the weird argot the author is laughing at his characters as much as laughing with them. It's a thing of its time that hasn't aged well.
Lazar said,
June 10, 2024 @ 4:38 pm
Extraneous asterisk in my comment above.
Andrew Usher said,
June 10, 2024 @ 7:56 pm
So you do agree, then, that that pronunciation, with /ˈə/ and not /ʌ/ , exists in American speech also, or even more? For 'gonna', of course, not 'because', though that might not be impossible either.
I hadn't even noticed about the asterisk, but as it is very evident that that contracted form is not present in standard American it could actually be justified, regardless of how it got there.
Lazar said,
June 11, 2024 @ 10:03 am
No, not really. I (American) use /ʌ/ in stressed gonna and because, and my impression is that most Americans do too; the /ˈə/ pronunciations have always struck me as a bit of a British distinctive (a sort of post facto half-imitation of our more extensive frum-rum split), although I'd imagine you could probably find some American who has it.
Lazar said,
June 11, 2024 @ 10:04 am
from-rum split, rather
Lazar said,
June 11, 2024 @ 10:14 am
from-rum merger! That's what I get for day drinking.
/df said,
June 13, 2024 @ 10:13 am
@Andrew Usher, thanks, you're quite right: the technical difference to which I was alluding is just cot/caught, /ɔ/ vs /o/, open-mid-back vs close-mid-back. But I stand by my entirely subjective account of tongue positioning and the effect on consonant production in the two cases for whatever that's worth.